fix: keep comments with statement in IndentNestedKeywords

[bhirsz]

Description

The IndentNestedKeywords formatter used to collect all comments from a run-keyword statement and move them in front of the formatted keyword. This caused two problems:

• Line-scoped disablers (e.g. # robocop: fmt: off) were relocated away from the line they were meant to affect.
• Comments were detached from the code they reference.

This PR makes the formatter keep comments together with their code:

• Trailing comments (sharing a line with data) are anchored to the closest keyword/argument and rendered on the same output line.
• Standalone comments (comment-only lines) are still kept as separate comment lines before the statement.

The same anchoring logic is now applied consistently across keyword calls and settings (Suite Setup/Suite Teardown/Test Setup/Test Teardown/[Setup]/[Teardown]), replacing the old "merge every comment onto the first line" behavior for settings.

Example

Before:

Run Keywords    Log    1    AND    Log    2    # keep me next to log 2

After:

Run Keywords
...    Log    1
...    AND
...    Log    2    # keep me next to log 2

Previously the comment was hoisted above Run Keywords.

Implementation

• misc.py: added split_comments_by_anchor() returning standalone comments plus a id(data_token) -> trailing comments map.
• IndentNestedKeywords.py: parse_keyword_lines appends trailing comments per output line; the visit methods use the anchor logic; node_was_formatted now strips comments from both sides before comparing.

Tests & docs

• Updated expected files and added a demonstrative case in the source data.
• All IndentNestedKeywords tests and the full formatter suite pass; ruff and mypy (strict) are clean.
• Documented the behavior in the class docstring and docs/formatter/formatters/IndentNestedKeywords.md.

Closes #1507